the grand divergence: global and european current account … · 2000. 8. 31. · kti/ie discussion...
TRANSCRIPT
-
MŰHELYTANULMÁNYOK DISCUSSION PAPERS
INSTITUTE OF ECONOMICS, CENTRE FOR ECONOMIC AND REGIONAL STUDIES,
HUNGARIAN ACADEMY OF SCIENCES BUDAPEST, 2015
MT-DP – 2015/42
The grand divergence: global and European
current account surpluses
ZSOLT DARVAS
-
2
Discussion papers
MT-DP – 2015/42
Institute of Economics, Centre for Economic and Regional Studies,
Hungarian Academy of Sciences
KTI/IE Discussion Papers are circulated to promote discussion and provoque comments.
Any references to discussion papers should clearly state that the paper is preliminary.
Materials published in this series may subject to further publication.
The grand divergence: global and European current account surpluses
Author:
Zsolt Darvas senior research fellow
Institute of Economics - Centre for Economic and Regional Studies Hungarian Academy of Sciences
senior fellow at Bruegel and at Corvinus University of Budapest
e-mail: [email protected]
August 2015
ISBN 978-615-5594-06-9
ISSN 1785 377X
-
3
The grand divergence: global and European current
account surpluses
Zsolt Darvas
Highlights
Global current account imbalances widened before the 2007/2008 crisis and have
narrowed since then. While the post-crisis adjustment of European current account
deficits was in line with global developments (though more forceful), European
current account surpluses defied global trends and increased.
We use panel econometric models to analyse the determinants of medium-term
current account balances. Our results confirm that higher fiscal balances, higher GDP
per capita, more rapidly aging populations, larger net foreign assets, larger oil rents
and better legal systems increase the medium-term current account balance, while a
larger growth differential and a higher old-age dependency ratio reduce it.
European current account surpluses became excessive during the past twelve years
according to our estimates, while they were in line with model predictions in the
preceding three decades.
Generally, the gap between the actual current account and its fitted value by the
model has a strong predictive power for future current account changes. Excess
deficits adjust more forcefully than excess surpluses. However, in the 2004-07 period,
excess imbalances were amplified, which was followed by a forceful correction in
2008-15, with the exception of European surpluses.
Keywords: Current account imbalances; Current account adjustment
JEL classification: F32, F41
Acknowledgement: The author is grateful to Bruegel colleagues and participants at the
CESifo-Delphi Conference on Current Account Adjustments for comments and suggestions,
to Menzie D. Chinn and Eswar S. Prasad for their suggestions for incorporating financial
development in the model, and to Pia Hüttl, Alvaro Leandro and Allison Mandra for research
assistance.
-
4
A nagy széttartás: globális és európai folyó
fizetésimérleg-többletek
Darvas Zsolt
Összefoglaló
A folyó fizetési mérlegek egyensúlyhiánya jelentősen emelkedett a 2007/2008-as válság
előtt és csökkent azóta. Amíg Európában a folyó fizetésimérleg-hiányok csökkentek a
globális folyamatokkal összhangban, addig az európai folyó fizetésimérleg-többletek
dacoltak a globális trendekkel és a tovább növekedtek. Tanulmányunkban panel
ökonometriai modellekkel elemezzük a folyó fizetési mérlegek középtávú meghatározóit.
Eredményeink megerősítik, hogy a magasabb költségvetési egyenlegek, a magasabb egy
főre jutó GDP, a gyorsabban öregedő népesség, a nagyobb nettó külföldi eszközök, a
nagyobb olajjövedelmek és a jobb jogrendszerek növelik a középtávú folyó fizetési mérleg
egyenlegét, míg a nagyobb növekedési eltérés és a magasabb időskori függőségi ráta
csökkenti. Az európai folyó fizetési mérlegek többlete – a becsléseink szerint – túlzottá
vált az elmúlt tizenkét évben, míg az ezt megelőző három évtizedben összhangban volt a
modellek eredményeivel. A tényleges folyó fizetési mérleg és a modell által becsült érték
közötti rés erős előrejelző képességgel rendelkezik a folyó fizetési mérleg jövőbeli
változására vonatkozóan. A túlzott folyó fizetésimérleg-hiányok erőteljesebben igazodnak,
mint a túlzott többletek. Azonban a 2004–2007 időszakban a túlzott fizetésimérleg-
egyensúlytalanságok felerősödtek, amelyet erőteljes korrekció követett 2008–2015 között,
kivéve az európai többleteket.
Tárgyszavak: folyó fizetésimérleg-egyensúlytalanságok; folyó fizetésimérleg-igazodás
JEL kód: F32, F41
-
1
1. Introduction
From the mid-1990s to the global economic and financial crisis, global current account
imbalances widened significantly. Figure 1 shows that the aggregate position of the world’s
57 ‘surplus countries’ increased from a surplus of 1 percent of their GDP in the mid-1990s to
about 7 percent by 2007, after which a steady decline started. The aggregate current account
position of 115 ‘deficit countries’ deteriorated to about -5 percent of their GDP by the crisis,
which was followed by a correction. Clearly, global current account imbalances have
significantly narrowed since 2008.
Figure 1
Global current account balances (% GDP), 1993-2015
-6
-4
-2
0
2
4
6
8
-6
-4
-2
0
2
4
6
8
1995 2000 2005 2010 2015
All countries
-16
-12
-8
-4
0
4
8
12
16
20
-16
-12
-8
-4
0
4
8
12
16
20
1995 2000 2005 2010 2015
Main oil producers
-6
-4
-2
0
2
4
6
8
-6
-4
-2
0
2
4
6
8
1995 2000 2005 2010 2015
Surplus countries Balanced countries Deficit countries
Non-EU non-oil countries (excl. US)
-8
-6
-4
-2
0
2
4
6
8
-8
-6
-4
-2
0
2
4
6
8
1995 2000 2005 2010 2015
European Union countries
Source: author’s calculations using the April 2015 IMF World Economic Outlook.
Note: we use the average current account balance in 2000-2007 to separate surplus (larger than 1%
of GDP), balanced (between 1% and -1% of GDP) and deficit (below -1% of GDP) countries. Thereby
we separate the 187 countries in our sample to 57 ‘surplus countries’, 15 ‘balanced countries’ and 115
‘deficit countries’. The three country sub-groups include 28 main oil producers (19 surplus, 3
balanced, 6 deficit), 28 European Union countries (8-2-18) and 130 non-EU non-oil countries
excluding the United States (30-10-90). We excluded the United States from the third panel, because
US current account developments are strongly influenced by the central role of the US dollar in the
international monetary system and due to its large size, as the US would dominate the aggregate of
non-EU non-oil deficit countries. Main oil producers are defined as oil rents more than 10% of GDP
on average. The current account balance is expressed in percent of the group GDP.
The correction of global imbalances was not just the result of smaller surpluses in the
main oil-exporting countries. The second panel of Figure 1 shows that the combined surplus
of the main oil producers reduced close to zero by 2015, so these countries surely played a
role. Yet the third panel, which reports the position of non-EU countries which are not
among the main oil producers, also shows a major decline in their surplus from about 6
-
2
percent of GDP in 2007 to about 2 percent in 2014, even if there is a slight expected increase
in 2015.
European Union current account surplus developments were different from the rest of the
world in recent years. There was only a small drop in the surplus from 6 percent of GDP in
2007 to about 5 percent in 2008-09, but since then a steady increase has started and the
expected surplus for 2015 is over 7 percent (panel 4 of Figure 1). While EU current account
deficits have forcefully corrected, the large and even increasing surpluses moved the EU’s
aggregate position from a broadly balanced position before the global economic and financial
crisis to a sizeable surplus. Thereby the EU became a major contributor to global current
account imbalances.
Euro-area surplus countries, and in particular the Netherlands and Germany, are the key
contributors to the EU’s current account surplus, yet Denmark (which maintains a fixed
exchange rate to the euro) and Sweden (which has a floating exchange rate) also report large
and persistent surpluses.
Different narratives can explain the increasing EU current account surplus. The
persistently high current surplus in a number of EU countries could be justified by various
fundamentals, such as the rapid aging of populations, which might require the accumulation
of savings. Another possible reason could be weak domestic demand and economic
developments (both in deficit and surplus countries), which temporarily depress imports
relative to exports. And regarding the adjustment of pre-crisis current account deficits, it is
arguable that they became ‘excessive’ before the crisis in a number of EU countries and these
deficits were bound to correct, thereby increasing the aggregate current account surplus of
the EU.
How important are these possible explanations for the increased current account surplus
of the euro area and the EU? Why do post-crisis EU current account surplus developments
differ so much from the developments in the surpluses of non-EU non-oil producer
countries? How large were ‘excess’ current account deficits inside and outside the EU? In
this paper, we answer these questions by estimating panel-econometric models to uncover
the medium-term determinants of current account balances.
2. Methodology
There is growing literature on estimating the medium-term determinants of current account
balances with panel econometric techniques. See for example, Chinn and Prasad (2003),
Gruber and Kamin (2007), Chinn and Ito (2007), Gagnon (2011), Lane and Milesi-Ferretti
-
3
(2012), Cheung, Furceri and Rusticelli (2013), Chinn, Eichengreen and Ito (2014) and
various IMF reports. These papers estimate the model:
(1) tik
tiktititi xxxCA ,)(
,
)2(
,2
)1(
,1, ,
where tiCA , is the current account balance (% of GDP) of country i in period t, )(
,
j
tix is the j-th
explanatory variable of country i in period t, j is the parameter of the j-th explanatory
variable and ti , is the error term. Most researchers estimate this model on 4 or 5-year long
non-overlapping sample periods to eliminate the impact of business cycles and use various
theories to motivate the explanatory variables. We do not add any country-specific or time
fixed effects, because we are interested in studying the impacts of the fundamental
determinants only.
The most frequently used explanatory variables are the following:
Fiscal balance (expected sign: positive): a deviation from Ricardian Equivalence will
imply that an increased fiscal deficit will lower national savings and thereby
deteriorate the current account balance. Similarly to Lane and Milesi-Ferretti (2012),
we measure fiscal balance relative to the weighted average of trading partners (as a
percent of GDP). In order to have a full sample from 1972 onwards, we are bound to
use only 41 trading parents for which data is available from 1972 as the reference
group.
Economic growth (expected sign: negative): faster economic growth can indicate
faster productivity growth, which could attract capital inflows and thereby worsen the
current account balance. We measure economic growth with real GDP growth
relative to the weighted average of 59 trading partners.
Stage of economic development (expected sign: positive): lower level of development
offers higher rate of return on capital according to neoclassical theory, which implies
that capital should flow from rich to poor countries, thus poor countries are expected
to have current account deficits. We measure the stage of economic development with
GDP per capita relative to the weighted average of 59 trading partners.
Various demographic variables were used in the literature:
o Young-age and old-age dependency ratios (expected sign: negative): the life-
cycle hypothesis suggests that young and old people save less, thus countries
with high young-age and old-age dependency ratios tend to have larger
current account deficits;
o Population growth (expected sign: negative): fast population growth might
suggest an increase in the share of young people, and thereby lower the
current account balance;
-
4
o Aging speed (expected sign: positive): countries in which the population is
getting old more rapidly should save more and thereby have a larger current
account surplus. This variable was introduced by Lane (2010) and
popularised by Lane and Milesi-Ferretti (2012) and is measured as the 20-
year forward-looking change in the old-age dependency ratio. For earlier
years we use the actual future change (e.g. for 1980 we use the actual change
from 1980 to 2000), while for more recent years we use the United Nations
2012 population projection (eg for 2005 we use expected change from 2005 to
2025, where the 2025 data is from the UN projection).
Oil rents as a percent of GDP (expected sign: positive): various indicators have been
used in the literature to isolate the impact of oil prices, production, consumption and
trade on current account balances. We use oil rents (percent of GDP), which is
influenced by oil price swings, as large oil rents typically lead increased exports which
are not matched by corresponding imports.
Net foreign assets as a percent of GDP (expected sign: positive): if the steady-state
NFA/GDP ratio is stable, in a growing economy a positive NFA position must be
accompanied by a positive current account balance. The NFA/GDP ratio is lagged (eg
the end-2011 value is used for the 2012-15 time period), as the NFA is determined by
the past current account balances (and valuation changes) and it provides an initial
condition for future current account balance.
Terms of trade (expected sign: positive): a change in world market prices of a
country’s exports relative to its imports is expected to improve the current account
balance.
Institutional quality (expected sign: positive): weak institutions lower the risk-
adjusted return on investment and thereby lead to lower capital inflows and
consequently lower current account balances. We proxy institutional quality with the
index of ‘Legal system and property rights’ from the Economic Freedom Network,
which is among the few indicators available for a sufficiently long period for a large
number of countries.
Financial development (expected sign: ambiguous): low level of financial
development might indicate an inefficient domestic financial system, which might
encourage savers to invest abroad, and thereby a low level of financial development
might coincide with a current account surplus. However, low level of financial
development could also indicate the presence of credit constraints, which lowers
private savings and thereby the current account surplus. We use two possible
indicators:
-
5
o The private credit/GDP ratio as a proxy for financial development, which is
the standard indicator used in the literature. It is imperfect, as it captures only
one aspect of financial development and might also signal the presence of
credit booms.
o The Financial Development Index (and some of its components) by Sahay et
al (2015), which is available only for 1980-2013 (and there are some missing
data for some countries for certain years).
Various dummy variables: some papers, such as Lane and Milesi-Ferretti (2012), use
various dummy variables: a crisis dummy to capture whether a country is
experiencing a major economic crisis; an Asian crisis dummy to capture the specific
disruptions of the countries concerned during the 1997/98 Asian crisis; a dummy for
financial centre to control for the possible measurement errors in the current account
of centres of international wholesale asset trade; and a dummy for Norway which is
interacted with net oil export to capture the country-specific institutional
arrangements that govern the management of Norway’s oil revenues. However, we
concluded that the determination of many of these dummy variables are
questionable, while some of these dummy variables did not prove to be statistically
significant in Lane and Milesi-Ferretti (2012) estimates. Furthermore, relative GDP
growth is included in our study, which can capture crisis situations. Therefore, we do
not include any dummy variable in our estimates.
After selecting the appropriate model, we will use the estimated model to calculate the
fitted current account values, which may correspond to a medium-term current account
‘equilibrium’ or ‘norm’. However, there are two issues suggesting that one should assess such
fitted values with caution.
First, our models might be imperfect and miss important variables – in which case the
fitted value might not correspond to an equilibrium notion. Yet as we will see, the estimated
gap between the actual and fitted current account has a strong predicting power for future
changes in the current account and for most countries the actual current account balance
fluctuates around the fitted value, which can be consistent with an equilibrium notion.
However, for a few countries like Australia or the United States, there are persistent gaps
between the actual and fitted values, suggesting that certain information for such countries is
missing from the model.
Second, we use the actual values of the explanatory variables to calculate fitted values for
the current account, but the actual explanatory variables do not always correspond to
medium-term sustainable values, even though we use time-averaged data over four-year
non-overlapping periods to eliminate fluctuations related to the business cycle. For example,
-
6
the actual fiscal position over a four-year period may not correspond to a medium-term
sustainable position.
The fitted values from our estimated models should be therefore assessed with caution,
yet the above-mentioned strong predictability result and the fluctuation of actual current
account balances around the predicted values for most countries suggests that our models
have useful informational content.
3. Data
In terms of the country-sample, we largely follow Lane and Milesi-Ferretti (2012), who
considered 67 countries, yet in their final regression 65 countries were used. Very small
countries and main oil producers are excluded, though Russia and Norway are included in
their sample. From these 67 countries we had to disregard Taiwan, because several variables
were not available, but added Malta to have all 28 European Union countries in our sample.
One variable is missing for Belarus, the index of legal system and property rights, and
therefore Belarus is not included in our models using this variable. Thus our models include
67 or 66 countries, depending on the use of index of legal system and property rights.
The time period we consider is 1972-2015, which we divide into eleven 4-year long non-
overlapping time periods. 2015 data are from IMF’s April 2015 World Economic Outlook
(WEO, for those variables which are included in this dataset). Our data sources are the
following:
Current account balance: the primary source is the IMF WEO; pre-1980 values and
some missing values were added from the IMF International Financial Statistics,
World Bank World Development Indicators and European Commission’s AMECO
database.
Fiscal balance: the primary source is the IMF WEO; pre-1980 values and some
missing values were added from Mauro et al (2013), European Commission’s
AMECO database and the EBRD’s Selected Economic Indicators database.
GDP growth: the primary source is the IMF WEO; pre-1980 values and some missing
values were added from World Bank World Development Indicators, European
Commission’s AMECO database, EBRD’s Selected Economic Indicators database and
Maddison Project.
GDP per capita at PPP: the primary source is the IMF WEO; pre-1980 values and
some missing values were chained backwards using data from World Bank World
Development Indicators, European Commission’s AMECO database, EBRD’s
Selected Economic Indicators database and the Maddison Project.
-
7
Young-age dependency ratio: World Bank World Development Indicators.
Old-age dependency ratio: World Bank World Development Indicators.
Population growth: World Bank World Development Indicators.
Aging speed (20-year forward-looking change in the old age dependency ratio):
calculated using data on old-age dependency ratio and United Nation’s population
projections (United Nations, Department of Economic and Social Affairs, Population
Division (2013). World Population Prospects: The 2012 Revision, DVD Edition.)
Oil rents (percent of GDP): World Bank World Development Indicators. Since the
most recent data point is 2013 and there were major oil prices changes since then, we
approximated 2014-2015 values by assuming that oil rents as a share of GDP evolved
proportionally with the evolution of oil prices.
Net foreign assets: the updated dataset of Lane and Milesi-Ferretti (2007).
Terms of trade: World Bank World Development Indicators and European
Commission’s AMECO database.
Index of ‘Legal system and property rights’: Economic Freedom Network.
Private credit/GDP ratio: World Bank World Development Indicators.
The Financial Development Index (and some of its components): Sahay et al (2015).
4. Medium-term determinants of current account balances: regression results
We start by replicating the two main models of Lane and Milesi-Ferretti (2012) in our
extended sample period, with the significant difference that we do not include any dummy
variable. Table 1 shows remarkable similarity between our and Lane and Milesi-Ferretti’s
results. Both the estimated values of the parameters and their significance are similar for the
fiscal balance, growth differential, GDP per capita, lagged NFA and the measure of oil1. For
aging speed, our parameter estimate is highly significant, while it was more insignificant in
Lane and Milesi-Ferretti (2012). Our parameter estimate for the dependency ratio is only
marginally significant (11 percent and 9 percent in the two models, respectively), while it was
significant in Lane and Milesi-Ferretti (2012). There are only two variables for which the
estimated sign of the parameter is not correct in our sample: population growth and the
terms of trade. The R2 of the regression is slightly lower in our estimation, which may be
explained by our disregard of the various dummy variables that Lane and Milesi-Ferretti
1 Lane and Milesi-Ferretti (2012) use ‘net oil balance (% of GDP)’, but we could not identify a data source for this variable and hence we use ‘oil rents (% of GDP)’, which is available in the World Bank’s World Development Indicators.
-
8
(2012) used, in addition to the differences in the sample, such as the time period (1969-2008
versus 1972-2015) and country coverage (65 versus 67).
Table 1
Medium term determinants of the current account balance:
replicating the Lane and Milesi-Ferretti (2012) models on our sample
Model without terms of trade Model with terms of trade
Lane and Milesi-Ferretti (2012)
This paper Lane and Milesi-Ferretti (2012)
This paper
Fiscal balance (+) 0.243*** 0.180*** 0.244*** 0.184***
(0.06) (0.051) (0.06) (0.051)
Growth differential (-) −0.072 -0.098 −0.08 -0.097
(0.09) (0.083) (0.09) (0.082)
GDP per capita (+) 0.027* 0.045*** 0.028* 0.044***
(0.01) (0.011) (0.02) (0.011)
Population growth (-) −0.74 0.147 −0.75 0.147
(0.47) (0.281) (0.48) (0.281)
Old dependency ratio (-) −0.15** -0.080 −0.16** -0.083*
(0.06) (0.05) (0.07) (0.05)
Aging speed (+) 0.056 0.171*** 0.046 0.176***
(0.06) (0.041) (0.06) (0.042)
Lagged NFA (+) 0.049*** 0.025*** 0.050*** 0.025***
(0.01) (0.008) (0.01) (0.008)
Oil balance (+) 0.239*** 0.239***
(0.06) (0.06)
Oil rents (+) 0.387*** 0.397***
(0.08) (0.078)
Oil balance Norway 0.14 0.171
(0.11) (0.13)
Log terms of trade (+) 0.0107 -1.381*
(0.01) (0.753)
Crisis dummy (+) 0.018** 0.018**
(0.01) (0.01)
Financial centre dummy 0.014 0.013
(0.01) (0.01)
Asian crisis dummy (+) 0.037*** 0.035**
(0.01) (0.01)
Observations 503 581 496 581
Time periods 10 11 10 11
Number of countries 65 67 65 67
R2 0.45 0.38 0.44 0.38 Note: Panel estimation, non-overlapping 4-year averages (except for the lagged NFA, which refers
to the last year of the previous 4-year period). The dependent variable is the average current
account balance during the 4-year period. The expected sign of the parameter is indicated in
brackets after the variable name. The Lane and Milesi-Ferretti (2012) sample include 10
observations between 1969-2008 for 65 countries, while our sample period includes 11 observations
between 1972-2015 for 67 countries. *,**, *** denote significance at 10, 5 and 1% levels respectively.
OLS estimation with robust standard errors.
-
9
Next, we extended the model with further variables discussed in the previous section:
young-age dependency ratio, private credit stock over GDP and the index of legal systems
and property rights. Separately, we also added the Financial Development Index of Sahay et
al (2015), which is available for a shorter sample period. We estimate the model for four
different country samples: EU, non-EU, advanced countries, emerging countries (see
Appendix A for the list of countries) in order to see whether the parameter estimates are
robust across different country samples.
Table 6 in Appendix B reports the detailed result for the full sample period (1972-2015) as
well as for the first (1972-1995) and second (1996-2015) part of the sample. The parameter
estimates of five variables are rather robust to alteration of the sample both in terms of
countries and time: budget balance, GDP per capita, aging speed, lagged net foreign asset
positon and oil rents. Three additional variables are estimated to have correctly signed and
mostly significant parameter in different samples: growth differential, old age dependency
ratio and the index of legal systems and property rights2.
We could not establish a robust relationship between current account developments and
domestic credit/GDP ratio (as a proxy for financial development): the parameter estimate is
practically zero (ie -0.001 with a 0.006 standard error) for the full sample, while in sub-
samples of different country groups the estimated parameter was significantly negative for
EU countries and advanced countries and significantly positive for emerging countries (and
non-significant for non-EU countries). The use of the Financial Development Index of Sahay
et al (2015) also suggests that results are rather different in different country groups. While
in the full sample of all countries the parameter estimate is significantly positive, this result
is driven entirely by emerging countries. For advanced countries and for EU countries the
parameter estimate is close to zero and not significant. Moreover, since the Financial
Development Index trends upwards for all countries, it is better to include it relative to
trading partners to capture whether financial development of a country in a given year was
higher or lower than in its trading partners. When we include the index this way, it was not
significant anymore in the full sample of all countries, its significance level dropped to 11
percent in the group of emerging countries, while it continued to be insignificant for EU and
advanced countries. Therefore, while we found some evidence for the importance of an
indicator capturing financial development for emerging countries, supporting the theory that
low level of financial development indicates the presence of credit constraints, which lowers
private savings and thereby the current account surplus (a finding similar to Chinn and
2 The index of legal systems and property rights is not available for Belarus and therefore the number of countries in our sample is reduced from 67 to 66.
-
10
Prasad, 2003), the estimated parameter is not significant for other country groups. For this
reason we did not include an indicator of financial development in our final model
specification.
The parameter estimates of population growth and young-age dependency ratio very
much depend on the time period and countries included in the sample and led in many cases
incorrectly signed and/or insignificant estimates. The parameter estimate of terms of trade
became consistently negative (which is an incorrect sign) and significant. We therefore
dropped these three variables from the model. Dropping these variables hardly changed the
parameter estimate of the other variables, suggesting that multicollinearity is not an issue.
Table 2 reports the regression results for our final model. For the sample of all countries,
parameters for six of the eight variables are highly statistically significant with correct signs,
while for the other two variables (growth differential and legal systems) the estimated sign is
correct, though the standard error is about the same as the parameter value. In some of the
country sub-samples the parameter estimate of these two variables is also significant. The
few cases with incorrectly estimated parameter signs are the following: growth differential
for the non-EU and advanced countries; the old-age dependency ratio for the EU and
advanced countries; and the index of legal systems and property rights for the non-EU
sample. When we restricted these parameters to zero, the estimated parameters of other
variables hardly changed, suggesting again that multicollinearity is not an issue.
Therefore, our results suggest that higher fiscal balance, higher GDP per capita, faster
aging speed, larger net foreign assets, larger oil rents and better legal systems increase the
medium-term current account balance, while a higher growth differential and larger old-age
dependency ratio reduce it.
The coefficient of determination (R2) suggests that the model fits the best for EU and
advanced countries (R2 is around 0.5), somewhat less for the non-EU sample (R2 = 0.41)
and less for the emerging country sample (R2 = 0.21). The R2 for the full sample is 0.38,
which suggests that the model explain a reasonably large share of the variation in current
account balances.
-
11
Table 2
Medium term determinants of the current account balance: our final model estimated for different country samples
All countries
EU EU
restricted Non-EU
Non-EU restricted
Advanced Advanced restricted
Emerging
Fiscal balance (+) 0.189*** 0.215*** 0.225*** 0.141** 0.129** 0.274*** 0.256*** 0.011
(0.05) (0.077) (0.076) (0.063) (0.063) (0.07) (0.074) (0.07)
Growth differential (-) -0.095 -0.311** -0.342** 0.061 0.164 -0.137
(0.085) (0.151) (0.141) (0.095) (0.185) (0.092)
GDP per capita (+) 0.041*** 0.085*** 0.086*** 0.033* 0.033** 0.077*** 0.076*** 0.034*
(0.012) (0.013) (0.013) (0.018) (0.015) (0.017) (0.017) (0.019)
Old dependency ratio (-) -0.094** 0.086 -0.074 -0.108 0.157*** -0.188***
(0.044) (0.071) (0.07) (0.069) (0.052) (0.061)
Aging speed (+) 0.16*** 0.151*** 0.172*** 0.172*** 0.179*** 0.142** 0.178*** 0.192***
(0.042) (0.05) (0.047) (0.065) (0.062) (0.056) (0.052) (0.059)
Lagged NFA (+) 0.025*** 0.007 0.006 0.03** 0.03*** 0.022** 0.02** 0.022***
(0.008) (0.007) (0.008) (0.012) (0.012) (0.009) (0.009) (0.007)
Oil rents (+) 0.387*** 0.355 0.398 0.393*** 0.414*** 0.006 0.056 0.415***
(0.081) (0.302) (0.304) (0.093) (0.09) (0.146) (0.15) (0.106)
Legal system (+) 0.113 0.719*** 0.726*** -0.107 0.37* 0.423** 0.164
(0.154) (0.235) (0.233) (0.209) (0.214) (0.215) (0.228)
Observations 570 224 224 346 355 277 277 293
Time periods 11 11 11 11 11 11 11 11
Number of countries 66 28 28 38 39 28 28 38
R2 0.38 0.50 0.50 0.41 0.41 0.51 0.49 0.21 Note: Panel estimation, non-overlapping 4-year averages (except for the lagged NFA, which refers to the last year of the previous 4-year
period) between 1972-2015. The dependent variable is the average current account balance during the 4-year period. The expected sign of the
parameter is indicated in brackets after the variable name.*,**, *** denote significance at 10, 5 and 1% levels respectively. OLS estimation
with robust standard errors.
-
12
5. Fitted current account balances
We use the estimated models reported in the previous section to calculate the fitted current
account values, which might correspond to the medium-term current account ‘equilibrium’
or ‘norm’ keeping in mind the caveats discussed in Section 2. Each country is included in
three country groups for which we estimated our model in Table 2: the global sample, either
EU or non-EU sample, and either the advanced country or the emerging country sample.
While Table 2 reports the quantitative differences between the estimated parameters along
these country-group samples, a plot of the fitted values is helpful for assessing the
differences across the models estimated for different country samples. Appendix C presents
the charts for all 66 countries included in our sample, while below we highlight the results
for three EU aggregate country groups and for a few specific countries inside and outside the
EU.
Figure 2
Actual and fitted current account balance (% GDP), 1972-2015: EU country
groups
-1
0
1
2
3
4
5
6
7
-1
0
1
2
3
4
5
6
7
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted from advanced model
EU surplus countries
-2
-1
0
1
2
3
-2
-1
0
1
2
3
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted from advanced model
EU balanced countries (FR&IT)
-8
-7
-6
-5
-4
-3
-2
-1
0
1
-8
-7
-6
-5
-4
-3
-2
-1
0
1
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted from emerging/advanced model
EU deficit countres (exlc. UK)
Note: Fitted values are derived from the estimation results (restricted versions) reported in Table 2.
The sample period includes 4-year non-overlapping periods (e.g. the last observation refers to 2012-
15). EU surplus (7): Austria, Belgium, Denmark, Finland, Germany, Netherlands and Sweden. EU
balanced (2): France and Italy. EU deficit (16): Bulgaria, Czech Republic, Cyprus, Estonia, Greece,
Hungary, Ireland, Latvia, Lithonia, Malta, Poland, Portugal, Romania, Slovakia, Slovenia and
Spain. Luxembourg and Croatia are not included in the aggregates due to their shorter time series,
while we left out the United Kingdom because its deficit developments very much differed from other
deficit countries. We use the average current account balance in 2000-2007 to separate surplus
(larger than 1% of GDP), balanced (between 1% and -1% of GDP) and deficit (below -1% of GDP)
countries. See the notes to Figure 1.
-
13
In order to highlight the general developments in EU surplus, balanced and deficit
countries (as we classified these countries for Figure 1), we calculated aggregates for the
three groups (Figure 2). There are some differences between the three fitted values from the
three models (in particular, the EU model tends to indicate somewhat larger fitted values),
but the dynamics are quite similar. For the ‘surplus countries’, the actual current account
fluctuated around the model estimates in 1980-2003, but since then large excess surpluses
emerged (note that our sample period includes 4-year averages and therefore we can only
highlight the start of that 4-year period when a major change is observed). The actual
position of the two ‘balanced countries’ (France and Italy) was quite similar to model
predictions, except in the 1990s, when both countries recorded excess surplus according to
our estimates, and in 2008-11, when they had a small excess deficit. In EU ‘deficit countries’
(not including the United Kingdom, which is shown separately in Figure 3), there were large
excessive deficits in 2000-2011, which were then rapidly corrected. On average, these
countries have not an excess surplus relative to model predictions of about 3 percent of GDP.
Let us assess the developments in some specific countries, which also allows considering
longer time periods.
Figure 3
Actual and fitted current account balance (% GDP), 1972-2015:
the three largest EU countries
-2
0
2
4
6
8
-2
0
2
4
6
8
72 80 88 96 04 12
Germany
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
72 80 88 96 04 12
Actual Fitted from global model Fitted from EU model Fitted from advanced model
France
-5
-4
-3
-2
-1
0
1
-5
-4
-3
-2
-1
0
1
72 80 88 96 04 12
United Kingdom
Note: Fitted values are derived from the estimation results (restricted versions) reported in Table 2.
The sample period includes 4-year non-overlapping periods (e.g. the last observation refers to 2012-
15).
Figure 3 reports the results for the three largest EU countries. For Germany, the actual
current account fluctuated around the fitted values in 1972-2003, but a persistent and large
positive gap emerged in 2004-15. In the latest time period, 2012-15, the ‘excess’ surplus of
-
14
Germany was about 5 percent of GDP according to the global model and about 3 percent
according to the EU and advanced country models. Interestingly, the fitted values for
Germany also increased recent years.
Table 3 decomposes the change in the fitted value between 2004-07 and 2012-15. For
Germany, the three main contributors are:
a. The relative fiscal balance (explaining 0.9 percentage point of GDP) increase: Figure
4 shows that while Germany’s fiscal position relative to its trading partners was close
to zero in 2004-07, its position in 2012-15 is about 4 percent of GDP higher. Whether
this improved relative fiscal position corresponds to long-term sustainability is an
open question. Yet as trading partners will most likely improve their fiscal position in
coming years, while it is not expected from Germany, Germany’s relative fiscal
position will likely decline and thereby reduce its estimated medium-term current
account equilibrium in the coming years;
b. Demographic factors (explaining 1.0 percentage point of GDP increase): the higher
share of old-age people implies a 0.3 percent of GDP lower current account balance,
while the rapid aging process implies a 1.3 percent higher balance;
c. The increased net foreign asset position (explaining 0.5 percentage point of GDP
increase): since Germany’s NFA position improved, more net income is expected,
which increases the current account balance.
Table 3
Contributions to the change in fitted current account values from
2004-07 to 2012-15 using the global model (% GDP)
Germany France
United Kingdom
(1) Fitted value 2004-2007 0.2 -0.4 -0.8
(2) Fitted value 2012-2015 2.6 -1.2 -1.4
(3)=(2)-(1) Change in fitted value 2.4 -0.8 -0.6
(3.1) Contribution of Fiscal balance 0.9 0.1 -0.2
(3.2) Contribution of Growth differential -0.1 0.0 -0.1
(3.3) Contribution of GDP per capita 0.2 -0.1 -0.1
(3.4) Contribution of Old dependency ratio -0.3 -0.3 -0.3
(3.5) Contribution of Aging speed 1.3 0.3 0.4
(3.6) Contribution of Lagged NFA 0.5 -0.8 -0.1
(3.7) Contribution of Oil rents 0.0 0.0 -0.1
(3.8) Contribution of Legal system -0.1 0.0 0.0
(4) Actual value 2004-2007 5.5 -0.4 -2.1
(5) Actual value 2012-2015 7.4 -1.0 -4.6
(6)=(4)-(1) Gap 2004-2007 5.2 0.0 -1.3
(7)=(5)-(2) Gap 2012-2015 4.8 0.2 -3.3 Note: the estimated model reported in the first column of Table 2 is used. The contribution of
each factor the change in the fitted value is the product of the estimated parameter and the
-
15
change in the variable from 2004-07 to 2012-15. The same decomposition for all countries is
reported in Appendix D.
Figure 4
The relative fiscal balance (% GDP), 1972-2015
-4
-3
-2
-1
0
1
2
3
4
5
-4
-3
-2
-1
0
1
2
3
4
5
72 76 80 84 88 92 96 00 04 08 12
Germany
France
United Kingdom
Note: The sample period includes 4-year non-
overlapping periods (e.g. the last observation refers to
2012-15).
France had relatively small actual deficits and surpluses in the past four decades and our
estimates reported in Figure 3 suggest that it had an ‘excess surplus’ in 1992-2003, which
corrected in later years. In the latest time period, 2012-15, France’s current account balance
was fully in line with the prediction of the global model, while it had a small (about 1 percent
of GDP) ‘excess’ deficit according to the EU model. The change in the fitted value from 2004-
07 to 2012-15 is entirely due to a worsened NFA position, as all other factors cancel out.
The actual balance of the United Kingdom fluctuated around the fitted values from 1972-
1999, after which a persistent negative gap emerged. In 2000-07 this gap was rather small
(around 1-2 percent of GDP depending on which estimated model we consider), but more
recently the gap widened and by 2012-15 our models suggest a 3-4 percent of GDP negative
gap.
-
16
Figure 5
Actual and fitted current account balance (% GDP), 1972-2015:
three euro-area deficit countries
-12
-10
-8
-6
-4
-2
0
2
-12
-10
-8
-6
-4
-2
0
2
72 80 88 96 04 12
Greece
-10
-8
-6
-4
-2
0
2
-10
-8
-6
-4
-2
0
2
72 80 88 96 04 12
Actual Fitted from global model Fitted from EU model Fitted from advanced model
Portugal
-10
-8
-6
-4
-2
0
2
-10
-8
-6
-4
-2
0
2
72 80 88 96 04 12
Spain
Note: Fitted values are derived from the estimation results (restricted versions) reported in Table 2.
The sample period includes 4-year non-overlapping periods (e.g. the last observation refers to 2012-
15).
The results for three euro-area deficit countries reported in Figure 5 show remarkable
similarity: before the mid-1990s (in the cases of Greece and Portugal) or late 1990s (in the
case of Spain), the actual current account balance fluctuated around the values predicted by
the model. Since then, up to the global financial and economic crisis, large excessive current
account deficits emerged, which adjusted quite abruptly, pushing the current account to a
small surplus, well over the values predicted by the models.
Figure 6
Actual and fitted current account balance (% GDP), 1972-2015:
three central European countries
-8
-6
-4
-2
0
2
4
-8
-6
-4
-2
0
2
4
72 80 88 96 04 12
Hungary
-20
-15
-10
-5
0
5
10
-20
-15
-10
-5
0
5
10
72 80 88 96 04 12
Actual Fitted from global model Fitted from EU model Fitted from emerging model
Latvia
-7
-6
-5
-4
-3
-2
-1
0
1
-7
-6
-5
-4
-3
-2
-1
0
1
72 80 88 96 04 12
Poland
-
17
Note: Fitted values are derived from the estimation results (restricted versions) reported in Table 2.
The sample period includes 4-year non-overlapping periods (e.g. the last observation refers to 2012-
15).
Central European member states that joined the EU in 2004 show similar patterns to the
three euro-area deficit countries discussed above (Figure 6). In the run-up to the crisis,
current accounts recorded larger deficits that what were predicted by the models, while the
most recent observations indicate positive gaps relative to the model predictions. It is
noteworthy that while for Hungary and Latvia the three models predict broadly similar
values, in the case of Poland the EU model predicts much larger current account deficits than
the predictions of the global and emerging country models.
We now turn to the assessment of the results for some non-EU countries.
Figure 7
Actual and fitted current account balance (% GDP), 1972-2015:
three non-EU advanced countries
-1
0
1
2
3
4
5
-1
0
1
2
3
4
5
72 80 88 96 04 12
Japan
-4
0
4
8
12
16
-4
0
4
8
12
16
72 80 88 96 04 12
Actual Fitted from global model Fitted from non-EU model Fitted from advanced model
Norway
-6
-5
-4
-3
-2
-1
0
1
2
3
-6
-5
-4
-3
-2
-1
0
1
2
3
72 80 88 96 04 12
United States
Note: Fitted values are derived from the estimation results (restricted versions) reported in Table 2.
The sample period includes 4-year non-overlapping periods (e.g. the last observation refers to 2012-
15).
Japan’s surplus was quite well in line with model predictions in 1980-99. After that, an
excess surplus emerged, which started to decline more recently, along with a decline in the
fitted values. The results for Norway are also interesting and suggest that the actual current
account was more or less in line with model predictions in 1972-99 and more recently in
2012-15, while in between the surplus was larger than what was predicted by the models. It is
noteworthy that the model predictions suggest a secular increase in Norway’s surplus from
the 1970s onwards. The United States is an outlier in the sense that the current account
balance has always been worse than what was predicted by the model. Most likely, the global
role of the US dollar and specific characteristics of US foreign assets and liabilities make it
-
18
possible for the US to run a current account deficit larger than what is predicted by a model
estimated on a large number of countries3.
Figure 8
Actual and fitted current account balance (% GDP), 1972-2015:
three non-EU emerging countries
-8
-6
-4
-2
0
2
-8
-6
-4
-2
0
2
72 80 88 96 04 12
Brazil
-4
-2
0
2
4
6
8
-4
-2
0
2
4
6
8
72 80 88 96 04 12
Actual Fitted from global model Fitted from non-EU model Fitted from emerging model
China, P.R.: Mainland
-8
-6
-4
-2
0
2
4
6
8
-8
-6
-4
-2
0
2
4
6
8
72 80 88 96 04 12
Korea
Note: Fitted values are derived from the estimation results (restricted versions) reported in Table 2.
The sample period includes 4-year non-overlapping periods (e.g. the last observation refers to 2012-
15).
Finally, Figure 8 shows our results for three main emerging economies. The actual current
account balance of Brazil fluctuated around the values predicted by our models, but China
had large surpluses well in excess of model predictions. More recently, the Chinese surplus
has declined, while the fitted value moved upwards (largely due to the rapid aging process,
and to a lesser extent due to improved fiscal position, increase in GDP per capita and
increased NFA – see Appendix D), thereby reducing the estimated excess surplus of the
country. Our results for Korea underline a secular increase in fitted values according to all
three models, while the actual current account fluctuated around the model predictions. The
1997 Asian crisis was followed by a moderate excess surplus (about 1.5-2 percent of GDP on
average in 1996-99), while in the next 12 years the actual current account was well in line
with model predictions. More recently, however, Korea’s current account surplus increased
to 6 percent of GDP, though our models predicted a smaller increase (largely due to the aging
process and to a lesser extent improved fiscal position and increase GDP per capita).
3 Gourinchas and Rey (2007) found for the United States that the cost of servicing its liabilities (which to a large extent comprise fixed income assets, partly reflecting the dominant role of the US dollar in the international monetary system) is much lower than the return on US investment abroad (which typically takes the form of various equity-type investments). Therefore, they have named the US the ‘World Venture Capitalist’.
-
19
6. Error correction
If the fitted values from our model correspond to an ‘equilibrium’ current account balance,
then we expect the actual current account to head for the predicted values. Therefore,
whenever there is an excess surplus, then either the actual surplus is expected to decline
towards the predicted surplus, or the predicted surplus is expected to increase towards the
actual surplus (or both). A simple test of the first chain of events is to estimate a regression
in which the change in the actual current account surplus is regressed on the previous period
gap between the actual and the predicted surplus, similar to an error correction model:
(2) tititi CAGAPCA ,1,, ,
where tiCA , is the change in the current account balance (% of GDP) of country i in period
t, 1,^
1,1, tititi CACACAGAP is the previous period gap between the actual values of
country i and the fitted values (% GDP), and ti , is the error term. A negative parameter for
would suggest that excessive imbalances are corrected. Lane and Milesi-Ferretti (2012, 2014)
estimated a variant of this regression concerning the change in the current account balance
from 2005-2008 either to 2010 or to 2012, using the current account gaps estimated in Lane
and Milesi-Ferretti (2012) for 2005-20084. They estimated a significantly negative value for
.
We estimated equation (2) for our full panel sample as well as for each time period as a
cross section regression. The results are reported in Table 4. The first block of the table
shows the results for our full panel sample. The estimated is significantly negative for all
country groups. The parameter estimate of -0.33 for the sample that includes all countries
indicate that one-third of excess current account surpluses and deficits are corrected on
average from one four-year period to the next, during our 40-year long sample period5. The
parameter estimates are somewhat higher in absolute terms for the EU (-0.55) and the
emerging country (-0.45), suggesting a stronger correction of current account gaps, while it
is slightly lower for the advanced country group (-0.25). The results clearly indicate that our
estimated current account gaps matter for the future development of the actual current
account, which is reassuring.
4 Milesi-Ferretti (2012, 2014) include other variables in the regression, like the lagged NFA position and a dummy for countries having fixed exchange rate regime. 5 Note that our full sample includes eleven 4-year long periods between 1972-2015, but since we use the lagged value of the current account gap, the effective sample period is reduced by one and thereby includes ten 4-year long periods between 1976-2015.
-
20
Table 4
Error correction regression results: symmetric specification of equation (2)
All EU non-EU Advanced Emerging
1972-2015 -0.33*** -0.55*** -0.31*** -0.25*** -0.45***
Full panel sample s.e. (0.05) (0.09) (0.06) (0.06) (0.07)
Nobs 504 196 308 249 255
R2 0.14 0.23 0.13 0.08 0.20
1976-79 -0.50*** -0.43 -0.45*** -0.59*** -0.21
cross section s.e. (0.07) (0.32) (0.09) (0.06) (0.15)
Nobs 27 11 16 18 9
R2 0.47 0.21 0.45 0.63 0.08
1980-83 -0.26* -0.2 -0.39** -0.1 -0.57**
cross section s.e. (0.13) (0.21) (0.18) (0.18) (0.2)
Nobs 33 13 20 20 13
R2 0.09 0.09 0.15 0.02 0.22
1984-87 -0.58*** -1.16** -0.61*** -0.57** -0.42**
cross section s.e. (0.13) (0.4) (0.17) (0.24) (0.17)
Nobs 41 14 27 24 17
R2 0.27 0.44 0.29 0.20 0.24
1988-91 -0.43*** -0.3 -0.49*** -0.32** -0.33
cross section s.e. (0.13) (0.25) (0.15) (0.14) (0.26)
Nobs 43 14 29 25 18
R2 0.20 0.14 0.22 0.15 0.10
1992-95 -0.18 -0.25 0.01 -0.17 -0.13
cross section s.e. (0.16) (0.2) (0.21) (0.18) (0.15)
Nobs 48 15 33 25 23
R2 0.03 0.07 0.00 0.03 0.03
1996-99 -0.21 -0.41 -0.33 0.06 -0.75**
cross section s.e. (0.20) (0.25) (0.29) (0.13) (0.32)
Nobs 53 19 34 26 27
R2 0.04 0.12 0.11 0.00 0.28
2000-04 -0.10 -0.28* -0.16 -0.18 -0.03
cross section s.e. (0.09) (0.16) (0.14) (0.13) (0.15)
Nobs 62 26 36 27 35
R2 0.02 0.13 0.03 0.05 0.00
2004-07 0.23** -0.27 0.13 0.48** 0.01
cross section s.e. (0.11) (0.22) (0.19) (0.19) (0.15)
Nobs 65 28 37 28 37
R2 0.06 0.04 0.02 0.23 0.00
2008-11 -0.47*** -0.67*** -0.33*** -0.30*** -0.67***
cross section s.e. (0.09) (0.20) (0.07) (0.06) (0.16)
Nobs 66 28 38 28 38
R2 0.48 0.41 0.41 0.45 0.55
2012-15 -0.54*** -0.58*** -0.46** -0.56** -0.45**
cross section s.e. (0.13) (0.17) (0.19) (0.21) (0.18)
Nobs 66 28 38 28 38
R2 0.25 0.27 0.25 0.21 0.20
Note: The first block reports result for the full panel sample, while the next 10 blocks show cross-
section results for each 4-year long time period. *,**, *** denote significance at 10, 5 and 1% levels
respectively. OLS estimation with robust standard errors.
-
21
The remaining ten blocks of the table report cross section results for each 4-year period.
For example, the second block of the table under the heading ‘1976-79’ reports the result of
the regression of the change in the current account balance from 1972-75 to 1976-79 as a
function of the 1972-75 current account gap. The parameter estimates are predominantly
negative: there are only 6 of the 50 estimates (10 time periods x 5 country groups) which lead
to a positive estimated parameter. Four of these 6 positive parameters are from the pre-crisis
period of 2004-07, suggesting that in the run-up to the crisis, instead of a correction of
existing current account imbalances, they have widened. This must have led to wider current
account imbalances by the crisis, which may explain why in the subsequent two periods,
2008-11 and 2012-15, the parameter estimates turn significantly negative again, while the
absolute value of parameter estimates are among the largest in these two periods. It is also
worthwhile that the R2 of this regression dropped close to zero in 2004-07, suggesting that
even though imbalances were amplified in this period according to parameter estimates,
existing imbalances explained very little of the variability of current account changes.
However, in the 2008-11 period, this simple regression explains about one half of the
variability, suggesting that the correction of imbalances during the crisis played a major role
in current account changes.
The results reported in Table 4 based on equation (2) assumed that the impacts of excess
surpluses and deficits are identical. Is this a correct assumption? In order to answer this
question, we allow the estimated parameter to differ whether there was an excess deficit or
an excess surplus in the previous period:
(3)
0
0
1,,1,22
1,,1,11
,
tititi
tititi
tiCAGAPifCAGAP
CAGAPifCAGAPCA
.
The results reported in Table 5 suggest that there is a major asymmetry: excess deficits
are more forcefully corrected than excess surpluses. Considering the full panel sample
reported in the first block of the table, the parameter estimates of excess deficit is highly
significant in all country groups, while the parameter estimates for excess surpluses are only
marginally significant. The absolute values of the estimated parameters are also much large
in the case of excess deficits. The cross-section results for the ten time periods reported in
the table also reveal that there are more cases with negative parameter estimates in the case
of excess deficits, and these parameters are more often significant than in the case of excess
surpluses.
Similarly to the symmetric case reported in Table 4, Table 5 also suggests that there some
‘error amplification’ in the pre-crisis period of 2004-07 (though parameter estimates are not
significant), while during and after the crisis, in 2008-11 and 2012-15, there were large and
-
22
statistically significant ‘error correction’ effects as regards excess deficits. The parameter
estimates for excess surpluses are correctly signed in 2008-11 and 2012-15, but statistically
significant only in the cases of non-EU and emerging countries in 2012-15. Therefore, our
results show that there was no statistically significant error correction in EU surplus
countries.
Table 5
Error correction regression results: asymmetric specification of equation (3)
All EU non-EU Advanced Emerging
1972-2015 1 -0.62*** -0.65*** -0.60*** -0.45*** -0.83***
Full panel sample s.e. (0.10) (0.19) (0.11) (0.12) (0.15)
2 -0.15 -0.09 -0.21* -0.05 -0.28**
s.e. (0.09) (0.14) (0.12) (0.11) (0.14)
Nobs 504 196 308 249 255
R2 0.17 0.25 0.16 0.11 0.24
1976-79 1 -0.59*** 1.18 -0.62*** -0.67*** -0.47
cross section s.e. (0.10) (0.75) (0.09) (0.07) (0.66)
2 -1.02** -0.52 -0.9*** -0.53 -0.07
s.e. (0.45) (0.30) (0.11) (0.32) (0.76)
Nobs 27 11 16 18 9
R2 0.53 0.40 0.76 0.64 0.12
1980-83 1 0.04 -0.20 -0.07 0.24 -0.89
cross section s.e. (0.22) (0.34) (0.50) (0.16) (0.87)
2 -0.13 -0.70*** 0.38* -0.20 0.52
s.e. (0.44) (0.20) (0.21) (0.58) (0.37)
Nobs 33 13 20 20 13
R2 0.13 0.33 0.29 0.11 0.31
1984-87 1 -0.68*** -1.38** -0.67*** -0.86*** -0.34
cross section s.e. (0.18) (0.62) (0.21) (0.21) (0.30)
2 -0.79** 1.88*** 1.22 0.13 -1.07
s.e. (0.35) (0.00) (1.39) (0.92) (0.64)
Nobs 41 14 27 24 17
R2 0.29 0.49 0.31 0.27 0.25
1988-91 1 -0.72* -0.88* -0.25 -0.46 -0.70
cross section s.e. (0.40) (0.40) (0.67) (0.42) (0.75)
2 -0.01 -0.08 0.04 0.09 -0.19
s.e. (0.18) (0.29) (0.26) (0.13) (0.20)
Nobs 43 14 29 25 18
R2 0.24 0.22 0.29 0.23 0.13
1992-95 1 -0.54 -0.41*** -0.07 -0.21 -0.02
cross section s.e. (0.40) (0.07) (0.60) (0.31) (1.01)
2 0.67* 1.72** 0.05 0.75 -0.38
s.e. (0.39) (0.64) (0.58) (0.69) (0.35)
Nobs 48 15 33 25 23
R2 0.14 0.68 0.00 0.29 0.05
1996-99 1 -0.65 -0.30 -0.63 0.12 -1.56***
-
23
cross section s.e. (0.58) (0.76) (0.59) (0.15) (0.53)
2 0.25 -0.28 0.29 -0.03 -0.5
s.e. (0.27) (0.27) (0.33) (0.42) (0.6)
Nobs 53 19 34 26 27
R2 0.11 0.13 0.21 0.01 0.39
2000-04 1 -0.18 -0.41 -1.12 0.33 -0.30
cross section s.e. (0.26) (0.37) (0.80) (0.77) (0.29)
2 -0.08 0.16 -0.24 -0.08 0.08
s.e. (0.16) (0.35) (0.18) (0.16) (0.41)
Nobs 62 26 36 27 35
R2 0.02 0.18 0.08 0.09 0.02
2004-07 1 0.01 0.50 0.18 0.08 -0.25
cross section s.e. (0.36) (0.43) (0.49) (0.47) (0.67)
2 -0.18 0.14 -0.14 0.09 0.08
s.e. (0.19) (0.72) (0.26) (0.37) (0.3)
Nobs 65 28 37 28 37
R2 0.16 0.14 0.09 0.31 0.01
2008-11 1 -0.67*** -0.80** -0.41*** -0.43*** -1.12***
cross section s.e. (0.20) (0.34) (0.08) (0.09) (0.28)
2 -0.12 -0.11 -0.07 -0.16 -0.04
s.e. (0.12) (0.33) (0.14) (0.19) (0.18)
Nobs 66 28 38 28 38
R2 0.54 0.43 0.49 0.48 0.69
2012-15 1 -0.86*** -0.64** -0.89* -1.40*** -0.23
cross section s.e. (0.28) (0.27) (0.47) (0.47) (0.28)
2 -0.47 -0.46 -0.58* -0.40 -0.81***
s.e. (0.32) (0.38) (0.29) (0.43) (0.26)
Nobs 66 28 38 28 38
R2 0.28 0.27 0.32 0.36 0.24 Note: The first block reports result for the full panel sample, while the next 10 blocks show cross-
section results for each 4-year long time period. *,**, *** denote significance at 10, 5 and 1%
levels respectively. OLS estimation with robust standard errors.
7. Conclusions
From the mid-1990s to the global economic and financial crisis, global current account
imbalances widened significantly, while there has been a major correction since then. The
adjustment of European current account deficits has been in line with global developments
(though they were more forceful), but European current account surpluses defied global
trends and continued to increase to over 7 percent of GDP. Thus, from a broadly balanced
current account position before the global crisis the EU became a major contributor to global
current account imbalances.
-
24
In order to assess various explanations for the EU’s increased surplus, we use a standard
panel econometric model to analyse the determinants of medium-term current account
balances. We estimate the model for 66 countries during eleven 4-year long non-overlapping
periods from 1972-2015, and study the model’s robustness in different time periods and
country samples. We consider several variables studied in previous literature and confirm
that higher fiscal balance, higher GDP per capita, more rapidly aging populations, larger net
foreign assets, larger oil rents and better legal systems increase the medium-term current
account balance, while a larger growth differential and a higher old-age dependency ratio
reduce it. We could not establish a robust relationship between current account
developments and the terms of trade, domestic credit/GDP ratio (as a proxy for financial
development), the financial development index of the IMF, population growth and the
young-age dependency ratio.
We found that in the first eight 4-year long periods of our sample, 1972-2003, the actual
balance of European surplus countries fluctuated around the predictions of our model, but in
the latest three 4-year periods, 2004-15, large positive gaps emerged. While worsening
demographic developments, improved fiscal positions and increased net foreign assets
explain some of the increase in European current account surpluses, they became excessive
according to our estimation results considering all variants of our model. Current account
deficits in several EU countries were highly excessive before the crisis according to our
results and were forcefully corrected. Most previous EU deficit countries display an excess
surplus now.
The gap between the actual current account balance and its fitted value in the model has a
strong predictive power for future current account developments in a panel specification that
treats surpluses and deficits symmetrically. An asymmetric model which allows different
correction of excess surpluses and deficits suggests that the adjustment of excess deficits is
more forceful than the adjustment of excess surpluses. Cross-section estimates for individual
4-year long periods suggests that the 2004-07 period was special in the past four decades,
because excess imbalances were amplified during this period, though such amplification
explained a small fraction of the variance of current account changes. In 2008-15, a forceful
correction followed, when the adjustment of earlier excess imbalances explained a large
share of the variation of current account balance changes, except for EU surplus countries.
Our key conclusion considering various versions of our models is that European current
account surpluses became excessive in the past twelve years as neither their level nor their
dynamics can be justified by standard panel econometric models. Such results raise
important policy questions and support the conclusions of the European Commission (2015)
and the IMF (2015) in their assessment of the current account surplus of Germany, the
country with the largest surplus in the EU. Decisive policy actions, such as bold structural
-
25
reforms and demand management, are needed to alleviate the problem of excessive current
account balances, as discussed by Darvas and Wolff (2014).
-
26
References
Cheung, Calista, Davide Fucieri and Elena Rusticelli (2013) ’Structural and cyclical factors
behind current account balances’, Review of International Economics 21(5), 923–944.
Chinn, Menzie D. and Eswar Prasad (2003) ’Medium-term determinants of current
accounts in industrial and developing countries’, Journal of International Economics
59, 47–76.
Chinn, Menzie D. and Hiro Ito (2007) ‘Current account balances, financial development
and institutions: Assaying the world "saving glut”’, Journal of International Money and
Finance 26(4), 546–569.
Chinn, Menzie D., Barry Eichengreen and Hiro Ito (2014) ‘A forensic analysis of global
imbalances’, Oxford Economic Papers 66(2), 465–490.
Darvas, Zsolt and Guntram B. Wolff (2014), ‘To the Commissioner for Economic and
Monetary Affairs’, In: Sapir, André (ed.): Memos to the new EU leadership – EU to do
in 2015-2019, Bruegel, p. 22-34, http://www.bruegel.org/publications/publication-
detail/publication/845-eu-to-do-2015-2019-memos-to-the-new-eu-leadership/
European Commission (2015) ‘Macroeconomic imbalances Country Report – Germany
2015’, European Economy, Occasional Papers 214,
http://ec.europa.eu/economy_finance/publications/occasional_paper/2015/pdf/ocp2
14_en.pdf
Gagnon, Joseph (2011) ‘Current account imbalances coming back’, Working Paper 11-1,
Peterson Institute for International Economics.
Gourinchas, Pierre-Olivier and Hélène Rey (2007) ‘From world banker to world venture
capitalist: US external adjustment and the exorbitant privilege’, in: Richard H. Clarida
(ed.) ‘G7 current account imbalances: sustainability and adjustment’, National Bureau
of Economic Research Conference Report, Chicago: University of Chicago Press, 11–66.
Gruber, Joseph W. and Steven B. Kamin (2007) ‘Explaining the global pattern of current
account imbalances’, Journal of International Money and Finance 26(4), 500–522.
IMF (2015) ‘2015 External sector report – Individual country assessments’,
http://www.imf.org/external/np/pp/eng/2015/062615a.pdf
Lane, Philip R. (2010) ‘International financial integration and Japanese economic
performance’, In: Kashyap, Anil, Hamada, Koichi, Weinstein, David (Eds.), Japan's
bubble, deflation and long-term stagnation, MIT Press, 129–174.
Lane, Philip R. and Gian Maria Milesi-Ferretti (2007) ‘The external wealth of nations
mark II: Revised and extended estimates of foreign assets and liabilities, 1970–2004’,
Journal of International Economics 73(2), 223–250.
Lane, Philip R. and Gian Maria Milesi-Ferretti (2012) ’External adjustment and the global
crisis’, Journal of International Economics 88, 252–265.
http://www.bruegel.org/publications/publication-detail/publication/845-eu-to-do-2015-2019-memos-to-the-new-eu-leadership/http://www.bruegel.org/publications/publication-detail/publication/845-eu-to-do-2015-2019-memos-to-the-new-eu-leadership/http://ec.europa.eu/economy_finance/publications/occasional_paper/2015/pdf/ocp214_en.pdfhttp://ec.europa.eu/economy_finance/publications/occasional_paper/2015/pdf/ocp214_en.pdfhttp://www.imf.org/external/np/pp/eng/2015/062615a.pdf
-
27
Lane, Philip R. and Gian Maria Milesi-Ferretti (2014) ’Global imbalances and external
adjustment after the crisis’, Working Paper 14/151, International Monetary Fund.
Mauro, Paolo, Rafael Romeu, Ariel Binder and Asad Zaman (2013) ‘A modern history of
fiscal prudence and profligacy’, Working Paper No. 13/5, International Monetary Fund.
Sahay, Ratna, Martin Cihak, Papa N'Diaye, Adolfo Barajas, Diana Ayala Pena, Ran Bi,
Yuan Gao, Annette Kyobe, Lam Nguyen, Christian Saborowski, Katsiaryna Svirydzenka
and Reza Yousefi (2015): ‘Rethinking Financial Deepening : Stability and Growth in
Emerging Markets Rethinking Financial Deepening: Stability and Growth in Emerging
Markets’, Staff Discussion Notes No. 15/8, International Monetary Fund
-
28
Appendix A: List of countries included in our sample
Note: we use the average current account balance in 2000-2007 to separate surplus (larger than 1%
of GDP), balanced (between 1% and -1% of GDP) and deficit (below -1% of GDP) countries. See the
notes to Figure 1.
EU Non-EU
Advanced
Emerging
Surplus
Balan
ce
Deficit
EU Non-EU
Advanced
Emerging
Surplus
Balan
ce
Deficit
Argentina x x x Latvia x x x
Australia x x x Lithuania x x x
Austria x x x Luxembourg x x x
Belgium x x x Malaysia x x x
Brazil x x x Matla x x x
Bulgaria x x x Mexico x x x
Canada x x x Morocco x x x
Chile x x x Netherlands x x x
China, P.R.: Mainland x x x New Zealand x x x
China,P.R.: Hong Kong x x x Norway x x x
Colombia x x x Pakistan x x x
Costa Rica x x x Peru x x x
Croatia x x x Philippines x x x
Cyprus x x x Poland x x x
Czech Republic x x x Portugal x x x
Denmark x x x Romania x x x
Dominican Republic x x x Russia x x x
El Salvador x x x Serbia x x x
Estonia x x x Singapore x x x
Finland x x x Slovakia x x x
France x x x Slovenia x x x
Germany x x x South Africa x x x
Greece x x x Spain x x x
Guatemala x x x Sri Lanka x x x
Hungary x x x Sweden x x x
Iceland x x x Switzerland x x x
India x x x Thailand x x x
Indonesia x x x Tunisia x x x
Ireland x x x Turkey x x x
Israel x x x Ukraine x x x
Italy x x x United Kingdom x x x
Japan x x x United States x x x
Korea x x x Uruguay x x x
-
29
Appendix B: Estimation results for the broad model
Table 6: Medium term determinants of the current account balance: the broad model estimated for different country samples
A: full sample 1972-2015
All countries
All countr. restricted
EU EU
restricted Non-EU
Non-EU restricted
Advanced Advanced restricted
Emerging Emerging restricted
Fiscal balance (+) 0.201*** 0.194*** 0.257*** 0.236*** 0.147** 0.083 0.277*** 0.242*** -0.003
(0.052) (0.052) (0.072) (0.076) (0.068) (0.071) (0.07) (0.065) (0.071)
Growth differential (-) -0.087 -0.104 -0.268* -0.375** 0.027 0.127 -0.111 -0.244***
(0.088) (0.087) (0.155) (0.149) (0.095) (0.194) (0.093) (0.084)
GDP per capita (+) 0.04*** 0.04*** 0.118*** 0.111*** 0.027 0.03* 0.089*** 0.087*** 0.051** 0.025
(0.013) (0.013) (0.014) (0.015) (0.018) (0.016) (0.017) (0.016) (0.021) (0.019)
Population growth (-) -0.021 -1.112** -0.594 0.356 -0.248 -0.539 0.088 -0.225
(0.422) (0.494) (0.458) (0.601) (0.646) (0.468) (0.568) (0.544)
Young depend. ratio (-) 0.021 0.195*** -0.06* -0.05** 0.066 -0.059* -0.055**
(0.029) (0.049) (0.034) (0.024) (0.066) (0.031) (0.024)
Old dependency ratio (-) -0.073 -0.096** 0.15** -0.123* -0.179** 0.186*** -0.23** -0.261***
(0.051) (0.045) (0.07) (0.071) (0.077) (0.066) (0.091) (0.09)
Aging speed (+) 0.196*** 0.168*** 0.363*** 0.243*** 0.088 0.097 0.291*** 0.276*** -0.038
(0.062) (0.051) (0.062) (0.05) (0.098) (0.096) (0.072) (0.065) (0.09)
Lagged NFA (+) 0.025*** 0.025*** 0.002 0.004 0.028** 0.029** 0.021** 0.02** 0.02*** 0.025***
(0.008) (0.008) (0.007) (0.007) (0.012) (0.012) (0.009) (0.008) (0.007) (0.007)
Oil rents (+) 0.403*** 0.389*** 0.474 0.457 0.408*** 0.44*** -0.052 0.447*** 0.311***
(0.084) (0.084) (0.307) (0.311) (0.088) (0.09) (0.164) (0.091) (0.096)
Log terms of trade (+) -1.481* -2.053 -1.922** 0.759 0.978 -3.084***
(0.797) (2.209) (0.8) (1.456) (1.275) (0.98)
Legal system (+) 0.175 0.162 0.829*** 0.831*** -0.361* 0.622*** 0.515** -0.333
(0.167) (0.157) (0.222) (0.236) (0.215) (0.23) (0.211) (0.222)
Private credit (+/-) -0.001 -0.001 -0.017*** -0.018*** 0.009 0.007 -0.02** -0.02*** 0.031*** 0.026***
(0.006) (0.006) (0.006) (0.006) (0.011) (0.01) (0.008) (0.007) (0.009) (0.008)
Observations 561 561 222 222 339 348 270 274 291 326
Time periods 11 11 11 11 11 11 11 11 11 11
Number of countries 66 66 28 28 38 39 28 28 38 39
R2 0.37 0.37 0.57 0.54 0.41 0.39 0.53 0.52 0.30 0.23
-
30
B: 1st part of the sample 1972-1995
All countries
All countr. restricted
EU EU
restricted Non-EU
Non-EU restricted
Advanced Advanced restricted
Emerging Emerging restricted
Fiscal balance (+) 0.125* 0.115* 0.042 0.04 0.213** 0.179* 0.101 0.085 -0.089
(0.064) (0.063) (0.087) (0.085) (0.096) (0.099) (0.075) (0.07) (0.109)
Growth differential (-) -0.092 -0.117 0.049 -0.098 -0.151 0.021 -0.034 -0.031 -0.18*
(0.091) (0.094) (0.196) (0.105) (0.11) (0.209) (0.216) (0.107) (0.093)
GDP per capita (+) 0.015 0.014 0.063** 0.061** 0.01 0.008 0.05*** 0.046*** 0.037 0.026
(0.013) (0.012) (0.025) (0.025) (0.02) (0.009) (0.018) (0.016) (0.026) (0.025)
Population growth (-) -0.102 -0.149 -1.584** -1.512*** 0.472 0.314 -1.184 -1.031
(0.701) (0.649) (0.647) (0.563) (0.928) (0.84) (1.069) (0.661)
Young depend. ratio (-) -0.006 -0.009 -0.028 -0.031 -0.007 -0.128 -0.138*** 0.021
(0.038) (0.033) (0.076) (0.075) (0.041) (0.078) (0.044) (0.049)
Old dependency ratio (-) 0.039 -0.196 -0.205* 0.122 0.114 -0.122 -0.128
(0.081) (0.125) (0.119) (0.122) (0.09) (0.143) (0.139)
Aging speed (+) 0.119* 0.116** 0.115 0.106 0.098 0.066 0.113 0.08 0.129
(0.066) (0.058) (0.137) (0.134) (0.092) (0.076) (0.081) (0.065) (0.256)
Lagged NFA (+) 0.019*** 0.02*** 0.018** 0.019*** 0.014* 0.016* 0.011 0.011 0.022*** 0.023***
(0.006) (0.006) (0.007) (0.006) (0.008) (0.008) (0.01) (0.008) (0.007) (0.007)
Oil rents (+) 0.141** 0.158** 0.346 0.331 0.143** 0.167** -0.029 0.103 0.053
(0.066) (0.067) (0.291) (0.299) (0.072) (0.072) (0.158) (0.064) (0.08)
Log terms of trade (+) -0.595 0.552 0.525 -1.337 0.776 0.898 -2.627**
(0.915) (2.323) (2.32) (1.023) (1.286) (1.067) (1.25)
Legal system (+) -0.209 0.465 0.472 -0.604*** -0.142 -0.337
(0.179) (0.302) (0.303) (0.224) (0.309) (0.265)
Private credit (+/-) 0.012 0.009 0.007 0.007 0.029** 0.02** 0.004 0.005 0.011 0.014
(0.008) (0.007) (0.011) (0.011) (0.012) (0.01) (0.01) (0.008) (0.015) (0.013)
Observations 242 248 85 85 157 161 136 142 106 135
Time periods 6 6 6 6 6 6 6 6 6 6
Number of countries 52 52 18 18 34 34 26 26 26 27
R2 0.28 0.28 0.42 0.42 0.30 0.27 0.40 0.42 0.19 0.11
-
31
C: 2nd part of the sample 1996-2015
All countries
All countr. restricted
EU EU
restricted Non-EU
Non-EU restricted
Advanced Advanced restricted
Emerging Emerging restricted
Fiscal balance (+) 0.233*** 0.242*** 0.475*** 0.448*** 0.064 0.086 0.725*** 0.521*** -0.002
(0.087) (0.087) (0.159) (0.149) (0.098) (0.099) (0.165) (0.127) (0.095)
Growth differential (-) -0.072 -0.156 -0.538** -0.627*** 0.239 -0.053 -0.062 -0.149
(0.147) (0.145) (0.251) (0.2) (0.166) (0.366) (0.154) (0.124)
GDP per capita (+) 0.064*** 0.059*** 0.13*** 0.12*** 0.071** 0.061* 0.108*** 0.096*** 0.076** 0.037
(0.02) (0.019) (0.022) (0.018) (0.035) (0.032) (0.023) (0.022) (0.03) (0.025)
Population growth (-) -0.298 -1.283** -1.134* -0.462 -0.193 -2.202** -1.355* 0.301
(0.554) (0.595) (0.59) (0.813) (0.78) (0.962) (0.773) (0.646)
Young depend. ratio (-) 0.099** 0.36*** -0.03 -0.03 0.364*** 0.015
(0.044) (0.115) (0.052) (0.04) (0.107) (0.053)
Old dependency ratio (-) -0.082 -0.165*** 0.146 -0.257** -0.266** 0.168 -0.138 -0.188***
(0.069) (0.059) (0.09) (0.104) (0.104) (0.11) (0.12) (0.066)
Aging speed (+) 0.196** 0.093 0.385*** 0.266*** -0.044 0.27** 0.173 -0.033
(0.085) (0.071) (0.089) (0.079) (0.139) (0.118) (0.111) (0.121)
Lagged NFA (+) 0.024** 0.024** -0.008 0.03* 0.031* 0.022** 0.018* 0.015 0.015
(0.01) (0.01) (0.009) (0.017) (0.016) (0.01) (0.01) (0.012) (0.012)
Oil rents (+) 0.662*** 0.551*** 0.472 0.518 0.595*** 0.552*** -0.428** 0.799*** 0.705***
(0.119) (0.111) (0.586) (0.614) (0.128) (0.132) (0.205) (0.113) (0.097)
Log terms of trade (+) -3.449** -2.457 -3.21** -0.661 -4.642***
(1.479) (6.389) (1.347) (4.076) (1.34)
Legal system (+) 0.356 0.31 0.473 0.977*** -0.117 0.071 -0.141 0.22 -0.122
(0.267) (0.265) (0.398) (0.361) (0.431) (0.393) (0.447) (0.43) (0.4)
Private credit (+/-) -0.006 -0.007 -0.023*** -0.021*** 0.005 0 -0.019 -0.024** 0.039*** 0.036***
(0.008) (0.008) (0.008) (0.008) (0.015) (0.012) (0.012) (0.01) (0.011) (0.008)
Observations 319 319 137 139 182 182 134 134 185 191
Time periods 5 5 5 5 5 5 5 5 5 5
Number of countries 66 66 28 28 38 38 28 28 38 39
R2 0.46 0.44 0.64 0.63 0.52 0.50 0.62 0.59 0.42 0.38
-
32
Appendix C: Actual and fitted current accounts for all countries in our sample
-4
-3
-2
-1
0
1
2
3
4
-4
-3
-2
-1
0
1
2
3
4
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted f rom advanced model
Austria
-3
-2
-1
0
1
2
3
4
5
6
-3
-2
-1
0
1
2
3
4
5
6
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted f rom advanced model
Belgium
-2
0
2
4
6
8
-2
0
2
4
6
8
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted f rom advanced model
Germany
-6
-4
-2
0
2
4
6
8
-6
-4
-2
0
2
4
6
8
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted f rom advanced model
Finland
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted f rom advanced model
France
-4
-3
-2
-1
0
1
2
3
-4
-3
-2
-1
0
1
2
3
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted f rom advanced model
Italy
0
2
4
6
8
10
12
14
0
2
4
6
8
10
12
14
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted f rom advanced model
Luxembourg
-2
0
2
4
6
8
10
12
-2
0
2
4
6
8
10
12
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted f rom advanced model
Netherlands
-8
-6
-4
-2
0
2
4
6
8
-8
-6
-4
-2
0
2
4
6
8
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted f rom advanced model
Matla
-4
-2
0
2
4
6
8
-4
-2
0
2
4
6
8
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted f rom advanced model
Denmark
-4
-2
0
2
4
6
8
-4
-2
0
2
4
6
8
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted f rom advanced model
Sweden
-5
-4
-3
-2
-1
0
1
-5
-4
-3
-2
-1
0
1
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted f rom advanced model
United Kingdom
-
33
-16
-12
-8
-4
0
4
-16
-12
-8
-4
0
4
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted f rom advanced model
Cyprus
-12
-10
-8
-6
-4
-2
0
2
-12
-10
-8
-6
-4
-2
0
2
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted f rom advanced model
Greece
-12
-10
-8
-6
-4
-2
0
2
4
6
-12
-10
-8
-6
-4
-2
0
2
4
6
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted f rom advanced model
Ireland
-10
-8
-6
-4
-2
0
2
-10
-8
-6
-4
-2
0
2
72 80 88 96 04 12
Actual
Fitted from global model
Fitted from EU model
Fitted f rom advanced model
Portugal
-10
-8
-6
-4
-2
0
2
-10
-8
-6
-4
-2
0
2
72 80 88 96 04 12
Actual
Fitted from g